The Money Overview

The IRS deployed 125 AI models to flag tax returns this year, up from 54 two years ago — crypto now triggers automatic matching

This spring marks the first time cryptocurrency brokers are required to report your sales directly to the IRS. If you sold Bitcoin, Ethereum, Solana, or any other digital asset in 2025, your exchange almost certainly filed a Form 1099-DA listing your gross proceeds. The IRS is now feeding those forms into the same automated matching systems it has used for decades to catch discrepancies on W-2s and traditional 1099s. The difference: crypto, an asset class that largely sat outside the agency’s information-reporting reach until this year, is now inside it.

At the same time, the agency’s artificial intelligence footprint has expanded sharply. The U.S. Treasury’s AI Use Case Inventory, updated in January 2026, lists roughly 125 AI-related applications across Treasury bureaus, including the IRS. That is more than double the 54 cataloged two years earlier. Together, the new data pipeline and the growing AI toolkit give the IRS a level of visibility into crypto transactions it has never had during a filing season.

How the IRS governs its AI systems

The IRS does not plug in algorithms on the fly. Under Internal Revenue Manual section 10.24.1, every model, algorithm, and dataset must be registered in an internal inventory before it goes into production. A companion section, IRM 10.5.1, governs how AI may be applied to tax administration specifically, covering uses like detecting potential noncompliance and fraud while prohibiting the use of taxpayer data to train public-facing AI tools.

Those internal rules sit within a broader legal framework. Statutory privacy protections under Title 5 of the U.S. Code limit how agencies collect, store, and repurpose personal information, including tax records. Criminal penalties for improper disclosure fall under Title 18. In practice, these guardrails constrain how aggressively the IRS can wield its expanding AI toolkit, even as the number of deployed models climbs.

Crypto reporting enters a new phase

The regulatory groundwork was laid in 2024. Final regulations published in the Federal Register defined which platforms qualify as brokers and which digital-asset dispositions must be reported. IRS Fact Sheet FS-2024-23 confirmed the timeline: brokers must report gross proceeds from sales on Form 1099-DA for transactions occurring on or after January 1, 2025, with cost-basis reporting phased in for certain transactions starting January 1, 2026.

The real shift in 2026 is the matching. The IRS TIN-matching program links taxpayer identification numbers on incoming 1099-DA filings to individual returns. When a broker reports that a taxpayer received $47,000 in proceeds from selling Solana, and that taxpayer’s Schedule D shows $32,000, the discrepancy can surface automatically, without a human examiner ever opening the file. To be clear, this is a hypothetical example, but it illustrates the mechanics. The process is identical to what the IRS has done with wage and investment income for decades. What changed is that crypto is now inside the system.

This did not happen overnight. The IRS added a digital-asset question to the front page of Form 1040 starting in 2019, and IRS Criminal Investigation has pursued crypto-related tax evasion for years, including issuing John Doe summonses to major exchanges. But those efforts relied heavily on targeted enforcement. Automated matching at scale is a fundamentally different capability.

How Treasury counts its AI applications

The 125 figure comes from Treasury’s January 2026 AI Use Case Inventory, published to comply with executive-order and OMB reporting mandates. The inventory covers the entire department, not just the IRS, and describes applications ranging from routine document routing to sophisticated risk scoring. Public descriptions remain high-level to avoid compromising enforcement operations.

Because the inventory catalogs “use cases” rather than individual production models, the count does not map one-to-one onto discrete algorithms running inside IRS systems. IRM 10.24.1 requires separate inventory entries for models, algorithms, and datasets, but the public-facing documents do not break out a standalone count for the IRS alone. What the inventory does establish is the scale of growth: Treasury’s AI footprint has more than doubled since 2024, when roughly 54 applications were cataloged, and the governance infrastructure, including mandatory registration and privacy reviews, has formalized to keep pace. For tax filers, the practical takeaway is that the IRS now has access to a substantially larger suite of automated tools than it did even two years ago, and those tools are being applied to an expanding pool of data that, for the first time, includes broker-reported crypto transactions.

What the IRS has not disclosed

Several important questions remain unanswered. The IRS has not revealed the thresholds its automated systems use before flagging a crypto-related discrepancy. It is unclear whether a $200 mismatch between a 1099-DA and a filed return will generate a notice, or whether the system prioritizes larger gaps that suggest deliberate underreporting. For taxpayers who traded across multiple platforms, transferred assets between wallets, or dealt with hard forks and airdrops, the potential for innocent mismatches is real.

The interaction between AI-driven risk scoring and traditional audit selection criteria is also opaque. Treasury’s inventory confirms that machine learning models help prioritize cases, but it does not reveal how heavily those scores weigh crypto activity compared with factors like income level, filing history, or prior noncompliance. A taxpayer who sold $5,000 worth of NFTs has no way of knowing whether that activity materially changes their audit odds.

Then there are practical capacity questions. The regulations set firm effective dates, but the pace at which the IRS can ingest, validate, and incorporate millions of new 1099-DA records into its matching systems depends on IT infrastructure that has historically lagged behind policy ambitions. If broker reporting is incomplete or inconsistent in the early years, the automated systems could produce a wave of erroneous notices. Alternatively, the agency could quietly delay full-scale matching while it works through data-quality issues. Neither scenario has been publicly addressed.

Practitioners in the tax and crypto space have noted the tension. As Shehan Chandrasekera, head of tax strategy at CoinTracker, has observed in public commentary, many taxpayers still do not realize that their exchange activity is now being reported to the IRS in the same way a brokerage account reports stock sales. That gap between taxpayer awareness and agency capability is, in his view, where the biggest compliance risks sit this filing season.

What crypto filers should do before the June 2026 deadline

For anyone who transacted in digital assets during 2025, the practical steps are straightforward even if the enforcement details are not. Gather transaction records from every platform and wallet used during the year. Compare those records against any 1099-DA forms received. Report all dispositions on Schedule D and Form 8949, including transactions where no 1099-DA was issued, because the obligation to report exists regardless of whether a broker filed paperwork. If numbers do not line up, attach a clear explanation rather than hoping the discrepancy goes unnoticed.

Tax practitioners are watching this filing season closely. The combination of new information returns and expanding AI capacity means the IRS will, for the first time, have a systematic way to identify crypto-related underreporting at scale. Whether the agency’s systems are fully operational or still ramping up, the regulatory framework is in place and the data is flowing. Filers who assume their crypto activity remains invisible to the IRS are making a bet that the numbers no longer support.

Avatar photo

Daniel Harper

Daniel is a finance writer covering personal finance topics including budgeting, credit, and beginner investing. He began his career contributing to his Substack, where he covered consumer finance trends and practical money topics for everyday readers. Since then, he has written for a range of personal finance blogs and fintech platforms, focusing on clear, straightforward content that helps readers make more informed financial decisions.​


More in IRS & Enforcement